I’m working on radar system.
We got new HW platform and we see very strong spikes in the time domain signal about 50db above the the target signal.
What can cause this phenomenon? I thought it might be a noise in the ADC that turning on a high bit...is it reasonable?
In case I can identify them, would it be fine if I will just zero these samples as they have low periodicity?
I think you will need to provide more details. Not being a radar person, I can make one comment -- it is probably not a good idea to zero-out data before you understand what is causing that data.
An example bad result from ignoring data is a metaphor for disaster: train wrecks!
Whenever I see such phenomena, the first thing I check is if I am reading the ADC bits correctly: Is it possible that you are reading a 16-bit word using the wrong byte order (big-endian vs little-endian)?
You said they have low periodicity, but is the phenomenon periodic? The period could give you some clues about the source (e.g. any relation to the sampling rate of the A/D?). Also worth investigating ambient interference that could be in your band of interest. Or maybe another processor, amplifier etc. on the same board.
Also, are the spikes of arbitrary magnitude or are they close to the saturation point of the A/D?
Zeroing out might end up being a practical solution ultimately, but I would definitely not do it before tracking down the source of the spikes.
Do you have the Tx functionning at the same time of the Rx?
Yes. It’s full duplex mechanism
Do you think it’s related?
Yes. Your Tx signal impinging on your Rx is very powerful compared to the signal coming back from the targets.
Ok. But I Dont see it in all Rx channels. Do you think it’s related to isolation issue between tx to rx
If you have check out all of the possible ADC issues below. You might want to check the hardware its self. You could be getting leakage from the transmitter in to the receiver. Is this a new design?
You definitely need to provide more details, and in so doing, may reveal to yourself what the problem may be. As suggested by others already:  maybe an ADC saturation issue, tx/rx isolation issue (are you using a RF circulator or signal clamp to suppress the Tx signal on the RX channel (which may be needed for HW protection?), just "ignoring" the input signal during tx may not be safe enough,  some analog amplifiers can suffer output polarity reversal during input saturation.  is the ADC uni- or bipolar? If unipolar, what protects the signal chain?  have you produced a model and used simulated data to "test" your algorithm?  have you used simulated signals to test your signal processing model's HW? These are all pieces of the overall system model and like any "chain" it's only as good as the weakest link.